| by VerifyID

Facial recognition – a learning science (Part 2 of 2)

Post image - Facial recognition – a learning science (Part 2 of 2)

Government-driven Global Competition

As we saw in Part 1 of this post, the availability of larger and larger datasets enabled development of progressively more powerful facial recognition systems.  In the mid-2000s, the US government began driving innovation in facial recognition with sponsored competitive evaluations and challenge problems that set the growing number of vendors against each other to test the success of their algorithms on a standardized set of scales. 

The National Institute of Science & Technology (NIST) initially ran the Facial Recognition Vendor Tests (FRVT) in 2000, 2002, and 2006 to measure progress of prototype systems/algorithms and commercial face recognition systems, then commenced an ongoing FRVT series that continues to this day.  NIST also sponsored the Face Recognition Grand Challenge from 2004 – 2006, which consisted of a series of increasingly difficult challenge problems using a dataset of 50,000 images. 

NIST continues to sponsor challenges of various sorts, publishing data on an ongoing basis, including several “Face Challenges” related to processing of unconstrained in-the-wild face images. 

More recently, the Department of Homeland Security (DHS) ran a Biometric Technology Rally in 2018 and has another currently underway for 2019.  These events are aimed at challenging the industry to develop more intuitive, faster, and more reliable collection systems.

VerifyID Intelligent Identity

The rapid evolution of both the technology for facial recognition and the algorithms that enable it can make products in this space obsolete before they hit the market. VerifyID’s engineers closely monitor challenge results and industry progress, and our modular design enables us to shift components as needed to ensure that we take advantage of the most accurate and reliable systems available. Furthermore, as our facial dataset grows into the millions of faces, accuracy continues to improve, bringing with it the promise of frictionless, trustful transactions in more and more venues and business models.